Currently, such deflects as misjudgment and inefficiency can be found in the technology of filtering Chinese abnormal mails. In order to efficiently solve this problem, this paper combines the advantages of random forest algorithm, adopts Chinese word segmentation method to extract features, and calculates the weight of word frequency. Based on a singular value degradation, this new approach performs better in filling in the algorithm to complete the detection of Chinese abnormal mail. Compared with the detection results of various algorithms, the experimental results show that the performance of the proposed random forest anomaly email detector is superior to other algorithms in accuracy, recall rate and time efficiency.